BOSTON UNIVERSITY COLLEGE OF ENGINEERING

Thesis

CHARACTERIZING ANKLE FUNCTION DURING STAIR ASCENT, DESCENT, AND LEVEL WALKING FOR ANKLE PROSTHESIS AND ORTHOSIS DESIGN

by DEANNA HELENE GATES B.S. University of Virginia, 2002

Submitted in partial fulfillment of the requirements for the degree of Master of Science 2004

Approved by

First Reader

______________________________________________________ Jim Collins, Ph.D. Professor of Biomedical Engineering Boston University, Boston, Massachusetts

Second Reader

______________________________________________________ Paolo Bonato, Ph.D. Director, Motion Analysis Laboratory Spaulding Rehabilitation Hospital, Boston, Massachusetts

Third Reader

______________________________________________________ Herbert Voigt, Ph.D. Professor of Biomedical Engineering Boston University, Boston, Massachusetts

Fourth Reader

______________________________________________________ Hugh Herr, Ph.D. Professor, MIT/Harvard Division of Health Sciences and Technology, Cambridge, Massachusetts

ii

ACKNOWLEDGEMENTS This thesis would not have been possible without the endless support I received in the last year. First, I would like to thank Paolo for taking me on as a student despite the lack of funding for this project. I really do appreciate it. Thank you to everyone in the Motion Analysis Lab at Spaulding. Jen, thank you for all your guidance with Vicon and taking the time to help me troubleshoot even when you had a million other things to be doing. Delsey, thank you for being so patient with me when I was learning how to use the accelerometers and to program in Matlab and never making me feel badly about asking so many questions. I really couldnt have done this without your support. And finally Mel, thanks for taking trips with me to AMTI and helping me with the stair set-up. I was truly lucky you took me on as your apprentice. I would also like to thank Hugh for providing the groundwork for this project. Your help with the modeling was also integral to this projects success. Thank you Ugo, for your help with the stair set-up and calibration. I couldnt have even begun testing without your help. Thank you to every one who agreed to take time from their research to come in and participate in the study. Finally, Id like to thank my parents for their endless support and encouraging me to pursue my interests even when that meant supporting me financially as well.

iii

CHARACTERIZING ANKLE FUNCTION DURING STAIR ASCENT, DESCENT, AND LEVEL WALKING FOR ANKLE PROSTHESIS AND ORTHOSIS DESIGN

DEANNA HELENE GATES ABSTRACT In this thesis, the ankle joint was studied during level walking, stair ascent, and stair descent to determine its biomechanics for use in the design of a new generation of prosthetic and orthotic systems. Data collection and analysis was performed at Spaulding Rehabilitation Hospitals Motion Analysis Laboratory. Ten healthy subjects were asked to walk across a walkway and walk up and down three steps. Marker and force plate data were recorded and processed to estimate kinematic and kinetic patterns of motion. The stance phase of level walking was broken down into sub-phases based on whether power was generated or absorbed at the ankle. Each sub-phase was examined to determine if it could be modeled using simple mechanical elements (springs, dampers, and torque actuators). Analysis of the relationship between ankle torque and ankle angular position and angular velocity led to results consistent with a previous study that showed that the stance phase of level walking can be modeled using linear springs, nonlinear springs and torque actuators [20]. To expand upon this study, ankle function was characterized during stair ascent and descent. During stair descent, the ankle function was separated into three sub-phases. During the first sub-phase, a significant amount of energy is absorbed at the ankle. Since, iv

the ankle velocity/torque plots did not suggest a linear relationship, a variable damper was suggested to model this phase. During the second phase, a greater amount of power was absorbed than was generated in the third phase. As such, a device that could store and release energy was explored. The two phases were effectively modeled using a linear spring with the addition of a variable damper to dissipate the excess energy. For stair ascent, ankle function was characterized in four phases; two energy storage and two energy release sub-phases. Both energy storage phases were modeled using linear

springs. Since the energy required to ascent stairs is so much greater than the energy absorbed, the remaining two phases required the use of a torque actuator to supplement the energy released by the spring. The final stage to characterizing ankle function was to study transitions between different ambulatory tasks. For the transition between level walking and stair ascent it was observed that the kinematics of initial stance is nearly identical to initial stance during level walking. During mid-stance, the ankle showed less power absorption than during level walking. During late stance, the ankle functioned in an identical fashion to stair ascent. Similar observations marked the results for the transition from stair descent to level walking. These results are promising to the design of ankle prosthetics and orthotics because they show that simple mechanical elements can be utilized and that it is possible to identify changes in ankle function so that the device can switch from one mode to the next.

TABLE OF CONTENTS ABSTRACT .......................................................................................................................... iv TABLE OF CONTENTS ........................................................................................................ vi LIST OF TABLES ............................................................................................................... viii LIST OF FIGURES ............................................................................................................... ix CHAPTER 1: INTRODUCTION .............................................................................................. 1 1.1 POPULATIONS BENEFITING FROM ANKLE FOOT PROSTHETICS AND ORTHOTICS....... 1 1.2 ANKLE FOOT PROSTHESIS AND ORTHOSIS DESIGN .................................................... 2 1.3 MODELS OF ANKLE BIOMECHANICS ......................................................................... 4 1.4 OVERVIEW OF THESIS MATERIAL .............................................................................. 6 CHAPTER 2: MATERIALS AND METHODS .......................................................................... 8 2.1 SUBJECT SELECTION .................................................................................................. 8 2.2 EQUIPMENT ................................................................................................................ 8 2.3 EXPERIMENTAL PROCEDURE .................................................................................... 10 2.4 MODELING ............................................................................................................... 11 2.5 CHARACTERIZING ANKLE FUNCTION ....................................................................... 12 CHAPTER 3: MODELING LEVEL WALKING ..................................................................... 15 3.1 THE GAIT CYCLE ..................................................................................................... 15 3.2 CONTROLLED PLANTARFLEXION ............................................................................. 18 3.3 CONTROLLED DORSIFLEXION .................................................................................. 20 3.4 POWERED PLANTARFLEXION ................................................................................... 22 CHAPTER 4: MODELING STAIR ASCENT AND DESCENT ................................................. 25 4.1 STAIR ASCENT ......................................................................................................... 27 4.2 CHARACTERIZING ANKLE FUNCTION DURING STAIR ASCENT ................................. 27 PHASE 1: CONTROLLED DORSIFLEXION 1 .................................................................... 29 PHASE 2: POWERED PLANTARFLEXION 1 ..................................................................... 31 vi

PHASE 3: CONTROLLED DORSIFLEXION 2 ........................................................................ 35

PHASE 4: POWERED PLANTARFLEXION 2 ..................................................................... 37 4.3 MODEL OF ANKLE DURING STAIR ASCENT .............................................................. 39 4.4 STAIR DESCENT ....................................................................................................... 40 4.5 CHARACTERIZING ANKLE FUNCTION DURING STAIR DESCENT ............................... 42 PHASE 1: CONTROLLED DORSIFLEXION 1 .................................................................... 42 PHASE 2: CONTROLLED DORSIFLEXION 2 .................................................................... 44 PHASE 3: POWERED PLANTARFLEXION ........................................................................ 49 4.6 MODEL OF THE ANKLE DURING STAIR DESCENT ..................................................... 55 CHAPTER 5. MODELING TRANSITIONS ............................................................................ 56 5.1 ASCENT TO LEVEL WALKING....................................................................................... 56 CHARACTERIZING THE TRANSITION FROM LEVEL WALKING TO STAIR ASCENT ..................... 61 5.2 DESCENT TO LEVEL WALKING ...................................................................................... 64 CHARACTERIZING THE TRANSITION FROM DESCENT TO LEVEL WALKING ............................. 66 CHAPTER 6. DISCUSSION .............................................................................................. 68 6.1 LOCATING TRANSITIONS .......................................................................................... 68 6.2 VALIDITY OF THE MODEL ........................................................................................ 69 6.3 SIGNIFICANCE OF THE WORK ................................................................................... 72 6.4 SUGGESTIONS FOR FUTURE WORK ........................................................................... 73 APPENDIX A. Graphs of ankle torque versus ankle angle and velocity for the CD1 phase of stair descent...74 APPENDIX B. Graphs of ankle torque versus angle and velocity for the transition from level walking to stair ascent .................................................................................. 76 APPENDIX C. Graphs of ankle torque versus angle and velocity for the transition from stair descent to level walking ................................................................................ 78 WORKS CITED .................................................................................................................. 80 CURRICULUM VITAE ........................................................................................................ 83 vii

LIST OF TABLES Table 2.1 Subjects sex, age and anthropometric data ...................................................... 8 Table 3.1 Summary of results of a linear regression of ankle torque on ankle angle for each phase of level walking for all subjects.............................................................. 21 Table 3.2 Work performed during each phase of level walking ...................................... 23 Table 4.1 Comparison of ankle angle at foot strike, and maximum and minimum angles and torques for level walking, stair ascent, and stair descent ................................... 26 Table 4.2 Results of a linear regression of ankle torque on ankle angular position during the first controlled dorsiflexion phase of stair ascent. .............................................. 31 Table 4.3 Work performed during each phase of stair ascent...32 Table 4.4 Results of a linear regression of ankle torque on ankle angular position during the first powered plantarflexion phase of stair ascent............................................... 33 Table 4.5 Results of a linear regression of ankle torque on ankle angle for the CD2 phase of stair ascent. ........................................................................................................... 35 Table 4.6 Results of a linear regression of ankle torque on ankle angle for the second powered plantarflexion phase of stair ascent. ........................................................... 39 Table 4.7 Work performed during each phase of stair descent........................................ 45 Table 4.8 Results for a linear regression of ankle torque on ankle angle for the CD2 and PP phase of stair descent........................................................................................... 48 Table 5.1 Parameter values for a linear regression of ankle torque on ankle angle for the transition from level walking to stair ascent ............................................................. 61 viii

LIST OF FIGURES Figure 2.1 The stairs manufactured by AMTI. .................................................................. 9 Figure 3.1 The gait cycle [26].......................................................................................... 16 Figure 3.2 Ankle angle, velocity, moment and power for level walking......................... 17 Figure 3.3 Results of a simple linear regression of ankle torque on ankle angle for the CP phase of level walking. ............................................................................................. 19 Figure 3.4 Results of a linear regression of ankle torque on ankle angle during the CD phase of gait .............................................................................................................. 20 Figure 3.5 Result of a linear regression of ankle torque on ankle angle for the PP phase of level walking......................................................................................................... 23 Figure 3.6 Schematic of the ankle model during level walking....................................... 24 Figure 4.1 The phases of stair climbing [25] ................................................................... 25 Figure 4.2 Ankle angle, velocity, moment and power for one cycle of stair ascent........ 28 Figure 4.3 Results of a linear regression of ankle torque on ankle angle for a single trial of subject KMC during the CD1 phase of stair ascent.............................................. 30 Figure 4.4 Results of a linear regression of ankle torque on ankle angular position during the PP1 phase of stair descent................................................................................... 34 Figure 4.5 Scatter plots for the CD2 phase of stair ascent............................................... 36 Figure 4.6 Plots of ankle torque on ankle angular position for the CD2 and PP2 phase of stair ascent................................................................................................................. 38 Figure 4.7 Schematic of the model of ankle function during stair ascent........................ 40

ix

Figure 4.8 Position, velocity, moment and power at the ankle for a single cycle of stair descent....................................................................................................................... 41 Figure 4.9 Plots of ankle torque versus ankle angle and velocity for the CD1 phase of stair descent............................................................................................................... 44 Figure 4.10 Results of a linear regression of ankle torque on ankle angular position during the CD2 phase of stair descent. ..................................................................... 47 Figure 4.11 Results of a linear regression of ankle torque on ankle angle during the powered plantarflexion phase of stair descent .......................................................... 50 Figure 4.12 Results of a linear regression of the residuals of the ankle angle and ankle torque regression on ankle angular velocity for powered plantarflexion ................. 51 Figure 4.13 Scatter plots of ankle torque versus ankle angle for a spring and damper model of the powered plantarflexion phase of stair descent..................................... 52 Figure 4.14 Scatter plot of ankle torque on ankle angle for the CD2 phase of stair descent....................................................................................................................... 54 Figure 4.15 Schematic diagram of the model of ankle behavior during stair descent..... 55 Figure 5.1 Ankle angle, velocity, torque and power for the transition from level walking to stair ascent............................................................................................................. 57 Figure 5.2 Scatter plot of ankle torque versus ankle angle for the transition from level walking to stair ascent............................................................................................... 59 Figure 5.3 Results of a simple linear regression of ankle torque on ankle angle for the transition from level walking to stair ascent. ............................................................ 60

Figure 5.4 Schematic of the models of ankle function during stair ascent, level walking and the transition from level walking to stair ascent. ............................................... 63 Figure 5.5 Graphs of ankle angle, velocity, torque, and power for the transition from stair descent to level walking............................................................................................ 65 Figure 5.6 Schematic of the ankle model during stair descent, level walking and the transition from stair descent to level walking ........................................................... 67

xi

CHAPTER 1: INTRODUCTION Ankle foot orthoses (AFOs) are devices used to improve the mobility of patients with numerous gait disabilities including cerebral palsy, stroke, and spinal cord injuries. These devices use some of the same technology as ankle prostheses for amputees. Currently available ankle prosthetic and orthotic systems are passive devices designed to improve mobility during level walking only. In a given day, a person completes many more tasks than this, as outside surroundings are by no means flat. In order for these systems to maximize the users mobility, they should adapt to the user and the tasks in which he/she engages. While the system should eventually adapt to any action, the focus of this research is on stair walking as stairs are frequently encountered in daily life, and require greater ranges of motion than those required by level walking [1].

1.1 POPULATIONS BENEFITING FROM ANKLE-FOOT PROSTHETICS AND ORTHOTICS The primary users of ankle prosthetic systems are above and below-knee amputees. Below-knee amputees account for about 50% of lower extremity amputees [2]. There are approximately 1.5 amputees per 1000 people in the U.S., or a total of 380,000 amputees [3]. A large number of amputations occur due to vascular disease (70%), such as diabetes, while other causes of amputation include trauma (23%), malignancy (4%) and congenital malformation (3%) [2]. A significant percentage of AFOs are used to help children with cerebral palsy. Cerebral palsy is one of the most common of all congenital disorders. Every year over 8,000 babies and infants are diagnosed with this condition [4]. This disorder can be 1

caused by a metabolic problem before birth, by an infection or stroke either before or after birth, or by a lack of oxygen reaching the fetus during delivery. Children with cerebral palsy have difficulties with balance, poor coordination, abnormal reflexes, and show a dramatic delay in developing motor skills. Orthotic intervention has increased in popularity over recent years to enhance mobility in children with cerebral palsy [4]. Another major population using AFOs is post-stroke patients. Stroke is a major cause of disability, with approximately 730,300 new cases in the United States annually, and about 160,000 Americans die each year from stroke (United States Agency for Health Care Policy and Research). AFOs are frequently prescribed for rehabilitation of hemiparetic patients suffering from drop foot or equinus foot. AFOs are recommended to minimize gait deviations caused by muscle tone changes or knee and ankle weakness [5]. The orthoses assist in foot clearance during the swing phase, improve the initial contact, and prevent ankle inversion injuries, as well as help to advance the body during midstance [6].

1.2 ANKLE-FOOT PROSTHESIS AND ORTHOSIS DESIGN There are two categories of prosthetic feet used by below-knee and above-knee amputees. The first category is the nonenergy-storing feet include the solidankle/cushioned-heel (SACH) foot and the single-axis foot. The SACH foot mimics ankle plantarflexion allowing for a smooth gait. The single-axis foot adds passive

plantarflexion and dorsiflexion, which increases stability during stance phase. Limitations in these non-energy storing prosthetic feet have led to the development of the 2

second category, dynamic elastic response designs, which incorporate modern, lightweight and elastic materials [7]. Energy-storing feet include, the multi-axis foot and the dynamic-response foot. Both feet add inversion, eversion, and rotation to

plantarflexion and dorsiflexion. To rate the efficacy of these new prosthetic designs, engineers use the following criteria: increased range of motion, better shock absorption and more energy recovery [7]. One key aspect of prosthetic design is the stiffness of the ankle-foot system. Selection of an appropriate stiffness is based upon the patients body weight, choice of activities, residual limb length and numerous other factors. Physicians must weigh the various factors and choose a single stiffness to serve all conditions a patient may experience [8]. Choosing the stiffness of the ankle-foot system inadequately may lead to increased metabolic costs, abnormal muscle activation patterns, and decreased gait symmetry. The ankle-foot system stiffness is designed to replace the functional behavior of the muscles and tendons of the intact biological limb. Unlike biological limbs, which can vary joint stiffness by altering the extent of muscle activation, the prosthesis is restrained to a single stiffness and is therefore not ideal for every situation [8]. Similar to ankle-foot prostheses, AFOs are also passive and therefore do not adapt to patient morphology and environmental disturbances. Clinicians can choose among several static and dynamic AFOs. Static orthoses prevent the motion of the ankle while dynamic AFOs allow some ankle motion. In the past, these devices have been

constructed using rigid thermoplastics lined with soft foam. More contemporary designs substitute these materials with lightweight and stiff composites. Additionally, current 3

fabrication methods are moving from craft intensive techniques to numerically controlled methods [9]. Despite progress in materials and methods however, all commercially available orthoses are passive devices.

1.3 MODELS OF ANKLE BIOMECHANICS The literature is replete with examples of how characterizing a biomechanical system by means of a model can be a tool for understanding the function of that system in a given situation. For the task of walking, the human leg has been characterized as a damped, linear, translational spring to understand its function during the stance phase of walking [10, 11]. Gilchrist and Winter also characterized the function of the human foot during stance using several independent damped spring systems [12]. Finally, the leg has also been characterized as linear, translational springs during bouncing gaits such as running, trotting, and hopping [13-18]. Modeling ankle function using simple mechanical devices is appealing to those attempting to design ankle-foot prostheses that replicate or mimic biology. Previous studies have shown that the ankle plantarflexors do not forcefully push the trunk upward and forward at the end of the stance phase of walking [19]. These studies imply that the non-disabled human ankle can be replaced effectively with a passive mechanical device. However, several other studies show that augmented systems may be necessary [19, 22]. Theses studies look at the moment at the ankle versus the ankle angular position to determine the stiffness of the ankle [19, 20]. The slope of the moment-angle curve is

viewed as the dynamic stiffness or quasi-stiffness of the ankle joint as it is not 4

measured at equilibrium [21]. Research looking at the moment versus ankle curves for level walking found the data created hysteresis loops, indicating that an augmented system would be necessary to replicate the ankle behavior [19]. In a 2002 Masters Thesis at MIT, Palmer studied the sagittal plane ankle moment versus ankle angle characteristics of level walking at different gait speeds. He modeled the stance phase of the gait cycle by separating it into three distinct sub-phases and found that augmented systems would be necessary at most walking speeds. This model focused on the sagittal plane because although the human ankle has the ability to move in all three planes, the vast majority of ankle work is done in the sagittal plane. The model therefore captures the most important features of ankle function during stance [22]. Recently, research has been done at MIT to design an active ankle-foot orthosis for the treatment of drop foot, a motor deficiency caused by total or partial paralysis of the anterior tibial muscle and peroneal group. This impairment causes slapping of the foot after heel strike and dragging of the foot during swing. To correct this deficiency at the ankle, Blaya and Herr developed an orthosis where the impedance of the joint is modulated throughout the gait cycle [23]. This orthosis utilizes a biomimetic torsional spring control to minimize forefoot collision with the ground and torsional spring damper control to lift the foot providing toe clearance. Their results showed that the orthosis was able to reduce the occurrence of slap-foot making the gait more biologically realistic [23]. This research shows promising results of what is possible with modeling and active control and could be expanded to a variety of tasks.

As walking plays an integral role in our daily lives, it has been extensively researched. Most of this research has been focused on level walking performed with the surety of a perfectly flat laboratory [24]. As research focuses around level walking, so do the models developed. This work will attempt to expand upon the models developed by Palmer to include stair walking and the transition between level walking and stair walking.

1.4 OVERVIEW OF THESIS MATERIAL The goal of this study was to develop a mechanical model to characterize the ankle function during the stance phase of stair ascent, descent, and the transitions between level walking and stair walking. To model each task, the stance phase was subdivided into sub-phases based on whether power was absorbed or generated. Using data obtained from ten healthy young adult subjects, correlations between ankle torque and ankle angle and velocity were explored. The ideal model was taken to be the simplest combination of mechanical elements (springs, dampers and torque actuators) that could reproduce the kinematic and kinetic patterns observed in the ankle data. Where simple mechanical elements were insufficient, devices such as variable dampers were suggested for additional definition in future work. Values of the model parameters (stiffness, damping constant) were found on a trial-by-trial basis and average values are reported. This thesis begins with a description of the methods used in the laboratory testing as well as the methods for data analysis and modeling. The third chapter reviews 6

Palmers model of the phases of level walking. Palmers analysis was replicated in this work and results for our ten subjects are given. In Chapter 4, stair ascent and descent are described. Results of regression analysis are given for each phase as well as the resulting model parameter values. Chapter 5 describes the transitions between level walking and stair ascent and stair descent and level walking. These tasks are also subdivided into phases and simple mechanical models were found to represent each phase. The final chapter gives a summary of all results and suggestions for future work.

CHAPTER 2: MATERIALS AND METHODS 2.1 SUBJECT SELECTION Ten healthy young adults participated in this research. The subjects had a mean age of 25 (range 23 to 29), a mean body mass of 63 kg (range 52.3 to 73.4), and a mean height of 170.6 cm (range 160 to 182.5). Subjects had no neurological, musculoskeletal or chronic ankle or knee problems. Prior to participation in the study, written informed consent was obtained from each subject.

Table 2.1 Subjects sex, age and anthropometric data Mean Leg Age Mass Height Subject Sex Length1 (yrs) (kg) (cm) (cm) SPC F 23 62.7 173 88.5 ELS F 28 66.5 161 86.3 HM M 26 52.3 182.5 98.3 CLH F 27 54.5 160 81 DJC M 24 67.8 172 89 NJR M 24 59.5 174 89 SEP F 23 62.2 166 83.3 MJK M 24 70.9 174 91 APG M 26 62.0 170 81 KMC M 29 73.4 173 89.5 Mean 25.4 63.2 170.6 87.7 SD 2.1 6.7 6.7 5.2 1 Leg length was measured from the medial malleolus to the ASIS

2.2 EQUIPMENT The lab is equipped with an 8 camera Vicon motion capture system (VICON 512, Oxford Metrics, Oxford, UK) and two AMTI force platforms (Watertown, MA). A stairway was designed and built by AMTI for use in this laboratory. Steps are connected 8

to two instrumented force platforms embedded in the walkway. The lowest step is attached to the second platform, the middle step to the first platform and the highest step to the second platform. The forces from each foot were measured independently as only one foot comes in contact with the force plate at a time during normal stair ambulation. The steps are offset so that the first foot contact is measured on the level force plate. The stairs are 7 inches high, 11.5 inches deep and 3 ft wide, in accordance with Massachusetts state building codes (See Figure 2.1). A short platform level with the highest step follows the stairs.

Figure 2.1: The stairs manufactured by AMTI. The steps are shown attached to two AMTI force platforms embedded in the walkway (platform following steps is not shown).

2.3 EXPERIMENTAL PROCEDURE Testing was completed at the Motion Analysis Laboratory at Spaulding Rehabilitation Hospital. 3 D pelvic and bilateral lower extremity joint kinematics and kinetics were collected via an 8-camera motion analysis system as the subject walked across a 10 meter level walkway with two staggered AMTI force platforms embedded in it. The camera system measured the

three-dimensional position of reflective markers, at 120 frames per second. Markers were attached to the pelvis (bilateral anterior superior iliac spines and posterior superior iliac spines), knee (lateral femoral condyles), ankle (lateral malleolus), forefoot (base of the second metatarsal), and heel. Additional markers were rigidly attached to wands over the mid-femur and mid-shaft of the tibia. Figure 2.2 shows the position of the reflective markers. The following Figure 2.2: Light reflective markers are placed over bony anatomical landmarks.

anthropometric measures were collected, along with motion analysis and force platform measures to calculate kinematics and kinetics: body weight, height, leg length (measured from medial malleolus to anterior superior iliac spine), knee width, and ankle width. In addition, six electromyographic active electrodes (Motion Lab Systems Inc, Baton Rouge, LA) were placed bilaterally on the gluteus medius, rectus femoris, medial head of biceps femoris, soleus, and tibialis anterior. All electrodes were attached to a box, which was attached to the subjects back. 10

Five repetitions including complete motion analysis data were collected and average lower extremity biomechanical values for each subject were obtained. For each trial, bilateral hip, knee, and ankle joint kinematic and kinetic data during walking in three planes (sagittal, coronal, and transverse) were collected. Ground reaction forces were measured synchronously with the kinematic data using the two staggered force platforms. Joint kinetics in each plane were calculated using a lower body inverse dynamic model. Data were normalized for body weight. Joint torque calculations were based on the mass and inertial characteristics of each lower extremity segment, the derived linear and angular velocities and accelerations of each lower extremity segment, as well as ground reaction force and joint center position estimates. Temporal parameters were obtained using the force platform and kinematic information to define foot contact times and distance parameters. These data were analyzed via a standard Vicon Plug-in Gait model (Oxford Metrics, Oxford, UK) to find the ankle moments. Moments for the segments during stair walking were calculated using a modified Vicon Plug-in Gait model to account for the stair height.

2.4 MODELING The data derived using a modified Body Builder model, including ankle angular position, and ankle torque, were then analyzed using MATLAB (Mathworks, Natick,MA).

11

The following sign conventions are used in this thesis: positive ankle position for dorsiflexion and positive ankle torque for dorsiflexion torque. The zero ankle position is defined as the point where the foot and shank are perpendicular. Ankle angular velocity was found by numerically differentiating the position. This position was filtered using a fifth order, zero-lag, low-pass Butterworth filter with a cut-off frequency of 6 Hz (Winter, 1990). Ankle power was estimated by multiplying the normalized ankle moments by the ankle angular velocity. Work was estimated by determining the area under the ankle moment versus ankle angle curve using a trapezoidal approximation [19]. While work could also be found by integrating power over time, the first method does not rely on numerical differentiation of the angular position data to obtain angular velocity and thus estimate power [19]. Thus estimation errors that accompany numerical differentiation do not affect the work estimate. This work is external to the ankle.

d Work = P dt = M dt = M d dt

(2.1)

2.5 CHARACTERIZING ANKLE FUNCTION Employing the same approach as Palmer, ankle function was characterized by considering ankle angular position and angular velocity as inputs into a black box and ankle torque as the output. With this idea, kinematic and kinetic data from 10 healthy subjects were used to find the simplest combination of mechanical elements to produce the observed position/torque and velocity/torque relationships. 12 These mechanical

elements represent the sum effect of the lengthening and shortening of the muscles, tendons, and ligaments as well as deformation of the foot and any other mechanisms of generating torque about the ankle joint [20]. In order to supplement the model of level walking described by Palmer, the same set of mechanical elements were considered for stair walking. These mechanical

elements include: torsional springs, torsional dampers, and torque actuators. Of these elements, springs and dampers are passive elements, while torque actuators are the only active elements. Torque actuators were considered for periods were the power at the ankle was positive and the amount of positive work done was greater than any energy that might have been stored by the ankle system [20]. Phases of stance during each activity were divided into periods based on whether power was generated or absorbed. If the direction of the ankle position or torque changed during any of these phases, the phase was broken down further. Since it would be more cumbersome to design a device that changed function due to power, changes in the position and velocity were used to identify the beginning and end of each phase. The category of passive devices was further differentiated between linear and non-linear elements. During periods where the ankle was characterized by spring-like behavior, the relationship between ankle torque and ankle angular position was expected to satisfy the equation

= k + 0

(2.2)

where is the ankle torque, k is the stiffness, is the ankle angular position, and o is the ankle torque when the position is zero. Values for these parameters were determined 13

using linear regression and the adequacy of the model was assessed by the R-square value and viewing scatter plots of the residuals versus the angular position. If the R-square value was low, or the residuals were highly patterned, the linear model was deemed insufficient [20]. During periods where the ankle was characterized by damper-like behavior, the relationship between ankle torque and ankle angular position was expected to satisfy the equation

= b

d + 0 dt

(2.3)

where is the ankle torque, b is the damping constant, d/dt is the ankle angular velocity, and o is the ankle torque when the ankle angular position is zero. Values for these parameters were found by performing a linear regression of the ankle torque versus ankle velocity curves. The adequacy of the model was determined in the same way as for spring-like behavior.

14

CHAPTER 3: MODELING LEVEL WALKING Given that level walking has been widely studied in the literature, the kinematics, and kinetics of the task are well defined. As such, it was important to look at level walking first in order to provide a basis of comparison and validate the testing procedure. In this portion of the work, subjects were asked to walk at a comfortable pace across a level walkway. Kinematic and kinetic data were collected and modeling of the ankle was performed to confirm the model found by Palmer was adequate for these subjects [20]. Data from this section were also used to compare parameter values for level walking and stair walking on an individual basis.

3.1 THE GAIT CYCLE In studies of human locomotion, a walking cycle is typically broken down into two phases, the stance phase (60%) and the swing phase (40%) (Figure 3.1). The gait cycle for the right side begins with heel strike of the right foot. At this point, both feet are on the ground. This is known as the initial double support phase. This sub-phase of the gait cycle is also known as weight acceptance as the body weight is shifted to one leg. Forward advancement begins when the left foot leaves the ground (ie. left toe-off). During the single support phase of stance, the right leg supports the body weight while the left leg advances forward. When the left foot hits the ground, it is the beginning of a second double support phase. As the right leg comes off the ground (toe-off), the body transitions into swing phase. During this phase, the limb advances forward in preparation for the next contact with the ground [25]. 15

Figure 3.1: The gait cycle [26] Characterization of the ankle during level walking was done in three parts, as described by Palmer [20]. The first period of stance was controlled plantarflexion (CP). This phase began at foot strike (FS) and ended at the point were the minimum ankle position was reached. This position was referred to as foot flat (FF). The second period of stance was controlled dorsiflexion (CD), which lasted from FF until the point where the power became positive. In this project, the end of controlled dorsiflexion was

detected by identifying the occurrence of the maximum value of angular position. This point corresponds to that chosen by Palmer (zero crossing of power trajectory) since the velocity at a maximum is zero, and power is the product of moment and velocity. The third period that was studied was powered plantarflexion (PP). This began the instant the power became positive and lasted until the foot came off the ground (FO). Foot strike 16

and foot off were located in the Vicon software using the ground reaction force with a force threshold of 20 N (Refer to Figure 3.2).

Controlled Plantarflexion

Powered Plantarflexion

Controlled Dorsiflexion

CD ends

FF

TO

Figure 3.2 Ankle angle, velocity, moment and power for a single trial of subject DJC for one gait cycle of level walking. 17

3.2 CONTROLLED PLANTARFLEXION During the controlled plantarflexion phase, Palmer characterized the ankle as a linear spring. This model was confirmed for the subjects tested with R-squared values between 0.74 and 0.96 (Table 3.1) and no clear pattern in the residuals. Results of a linear regression of ankle torque on ankle angle are shown for a representative subject in Figure 3.3. One reason the fit may not have been strong in a few of the subjects is that any noise in the kinematic or kinetic data could cause a significant effect in the fitting since the phase is so short. For instance, subject SPC had the lowest average R-squared value of 0.74. For this subject, the moment decreased slightly in the middle of the phase. This change may have been noise. Since the duration of the phase was approximately 78 ms., which at a sampling rate of 120 Hz means the fitting used 9 points, even just one outlier, causes a significant decrease in R-squared value. Since the results for the majority of subjects confirmed the strong linear relationship found by Palmer, it was assumed that the model was adequate for all subjects.

18

Figure 3.3 Results of a simple linear regression of ankle torque on ankle angle for the CP phase of level walking. Data are shown for a single trial of subject DJC, R2=0.96.

19

3.3 CONTROLLED DORSIFLEXION During controlled dorsiflexion, Palmer characterized the ankle as a non-linear spring that gets stiffer as it is stretched. Results of the analysis performed in this project support this assumption. The R-squared values between ankle torque and ankle angle were between 0.78 and 0.96 while the residuals for all cases were highly patterned, indicating that a linear spring would not suffice. Results of this analysis are shown in Figure 3.4 and Table 3.1.

Foot Flat

CD ends

Figure 3.4 Results of a linear regression of ankle torque on ankle angle during the CD phase of gait. Data are shown for the same trial as Figure 3.3 (R2=0.92).

20

Table 3.1 Summary of results of a linear regression of ankle torque on ankle angle for each phase of level walking for all subjects. Spring Stiffness # of Points Subject Time in phase # of trials Phase k R2 Used for Label (sec) (Nm/kg-deg) Regression CP 0.078 (0.005) -0.020 (0.001) 0.74 (0.07) 9 (0.5) SPC 5 CD -0.085 (0.007) 0.84 (0.05) 58 (1.6) 0.470 (0.011) PP -0.070 (0.006) 0.99 (0.00) 15 (0.6) 0.125 (0.005) CP -0.028 (0.007) 0.89 (0.02) 9 (0.8) 0.079 (0.006) ELS 8 CD -0.052 (0.005) 0.96 (0.02) 59 (1.8) 0.465 (0.015) PP -0.050 (0.005) 0.99 (0.01) 12 (1.1) 0.104 (0.001) CP 0.029 (0.068) 0.78 (0.04) 8 (0.8) 0.070 (0.007) HM 5 CD -0.010 (0.004) 0.88 (0.03) 58 (4.2) 0.477 (0.030) PP -0.077 (0.006) 0.99 (0.00) 15 (2.0) 0.122 (0.011) CP -0.001 (0.006) 0.81 (0.10) 9 (1.1) 0.073 (0.009) CLH 5 CD -0.046 (0.006) 0.85 (0.06) 57 (2.9) 0.477 (0.025) PP -0.055 (0.008) 0.98 (0.01) 12 (1.3) 0.098 (0.010) CP -0.078 (0.014) 0.94 (0.02) 7 (1.0) 0.058 (0.008) DJC 7 CD -0.086(0.008) 0.89 (0.06) 57 (1.7) 0.479 (0.014) PP -0.068 (0.006) 0.99 (0.00) 14 (0.7) 0.118 (0.006) CP -0.026 (0.006) 0.84 (0.08) 11 (1.0) 0.087 (0.009) NJR 6 CD -0.072 (0.015) 0.89 (0.04) 61 (3.0) 0.511 (0.025) PP -0.081 (0.008) 0.99 (0.00) 15 (2.2) 0.128 (0.018) CP -0.026 (0.014) 0.85 (0.09) 7 (2.2) 0.060 (0.018) SEP 5 CD -0.092 (0.007) 0.89 (0.05) 55 (2.2) 0.458 (0.018) PP -0.055 (0.009) 0.99 (0.00) 18 (1.4) 0.150 (0.012) CP -0.066 (0.019) 0.92 (0.02) 6 (0.9) 0.047 (0.007) MJK 7 CD -0.094 (0.024) 0.89 (0.06) 58 (2.1) 0.483 (0.018) PP -0.069 (0.004) 0.99 (0.00) 16 (1.3) 0.130 (0.011) CP -0.067 (0.027) 0.94 (0.05) 5 (1.0) 0.043 (0.008) AG 6 CD -0.085 (0.008) 0.78 (0.04) 56 (1.5) 0.465 (0.012) PP -0.068 (0.009) 0.99 (0.00) 13 (2.4) 0.113 (0.020) CP -0.038 (0.009) 0.96 (0.02) 9 (1.1) 0.078 (0.009) KMC 8 CD -0.090 (0.010) 0.89 (0.1) 66 (1.2) 0.550 (0.010) PP -0.054 (0.008) 0.97 (0.01) 16 (1.4) 0.136 (0.012)

21

3.4 POWERED PLANTARFLEXION During powered plantarflexion, the ankle was characterized using a torque actuator that assists the non-linear spring during foot plantarflexion. Palmer found that the amount of work done by the torque actuator increased with gait speed. Since, in this study subjects were asked to walk at a comfortable pace, the results were expected to resemble those reported by Palmer for normal speed. During powered plantarflexion, the amount of work generated at the ankle was greater than that absorbed for eight of the ten subjects (Table 3.2). For those subjects where the work generated during powered plantarflexion was less than that absorbed, the average walking speed was lower than for the other subjects. As Palmer showed, when the gait speed decreases, the amount of work that the torque actuator does decreases. In this sense, the results are in agreement with those found previously. Results of a linear regression of ankle torque on ankle angle are shown in Figure 3.5. Regression analysis shows that the relationship between ankle torque and ankle angle is linear. The model derived for ankle biomechanics during level walking is summarized in Figure 3.6. The schematic representation shown in this figure demonstrates a state machine corresponding to the controller of a prosthetic or orthotic device assisting ankle biomechanics during ambulation. For the sake of completeness, a state corresponding to the swing phase is also shown. Although the implementation of the swing-phase model is not described in this thesis, it is envisioned that prosthetic and orthotic devices designed by using the approach herein proposed could rely on a second-order, underdamped mechanical model as per previous work by Blaya and Herr [23]. 22

Table 3.2. Summary of the work generated during each phase of level walking Average Average Average WPP Subject # of WCD WPP W CP (|WCP|+| WCD|) Label trials (J / kg) (J / kg) (J / kg) (J / kg) SPC 6 -0.37 -7.42 13.15 5.36 ELS 8 -0.83 -14.46 14.94 -0.34 HM 5 -0.12 -13.25 25.19 11.82 CLH 5 -0.32 -12.44 11.93 -0.84 DJC 7 -0.25 -12.66 19.63 6.72 NJR 6 -0.83 -9.29 12.69 2.56 SEP 5 -0.13 -9.85 22.93 12.95 MJK 7 -0.13 -7.81 17.21 9.26 AG 6 -0.27 -6.09 16.73 10.37 KMC 8 -0.32 -10.30 16.26 5.64

Figure 3.5 Result of a linear regression of ankle torque on ankle angle for the PP phase of level walking. Data are shown for the same trials as Figures 3.3-3.4 (R2 =0.9985). 23

Figure 3.6 Schematic of the ankle model during level walking. The CP phase is modeled as a linear spring, the CD phase as a non-linear spring, and the PP phase as a torque actuator that assists the non-linear spring during ankle plantarflexion.

CD1 CP

SW

CD CD2

PP
T

24

CHAPTER 4: MODELING STAIR ASCENT AND DESCENT Stair climbing is a common activity of daily living. This action is more stressful than level walking since the body must be lifted with each ascending step and continuously decelerated with each descending step [27]. Stair negotiation has three main requirements. The first is generation of concentric forces to propel the body upstairs or eccentric forces to control the bodys descent. Second, individuals must be able to control their center of mass within a constantly changing base of support. Finally, individuals must also have the capacity to adapt strategies to accommodate changes in stair environment, such as height, width, and the presence or absence of railings [28]. Like level walking, stair climbing has been divided into two phases (shown in Figure 4.1), a stance phase lasting approximately 64% of the full cycle and a swing phase lasting approximately 36% of the cycle [29].

Figure 4.1 The phases of stair climbing [25] 25

Table 4.1 Comparison of ankle angle at foot strike, and maximum and minimum angles and torques for level walking (LW), stair ascent, and stair descent Min Angle at Max Dorsi. Max Dorsi. Min Plantar Subject #of Plantar Condition Contact Angle Torque Torque Label trials Angle (deg) (deg) (Nm/kg) (Nm/kg) (deg) LW 6 -3.0 (0.9) 7.8 (0.9) -16.4 (2.6) 0.13 (0.02) -1.39 (0.08) SPC Ascent 5 13.3 (2.6) 17.5 (1.2) -19.3 (5.6) 0.01 (0.00) -1.57 (0.72) Descent 4 -16.8 (5.3) 21.7 (1.4) -26.6 (0.9) 0.08 (0.07) -1.36 (0.14) LW 8 -4.1 (1.7) 18.1 (1.2) -21.6 (6.8) 0.20 (0.03) -1.27 (0.10) ELS Ascent 5 10.0 (2.7) 16.1 (1.1) -31.4 (8.1) 0.03 (0.01) -1.33 (0.13) Descent 6 -27.7 (8.6) 32.0 (7.4) -36.0 (5.2) 0.02 (0.01) -1.57 (0.24) LW 5 -2.1 (2.3) 13.6 (1.4) -22.8 (5.6) 0.09 (0.02) -1.94 (0.14) HM Ascent 6 16.6 (3.2) 21.6 (2.3) -18.5 (3.1) 0.07 (0.06) -1.79 (0.19) Descent 5 -24.9 (2.5) 33.4 (1.0) -28.3 (1.8) 0.02 (0.00) -1.49 (0.25) LW 5 -5.2 (2.1) 16.0 (2.9) -17.4 (4.7) 0.08 (0.01) -1.29 (0.02) CLH Ascent 4 14.0 (2.3) 18.0 (1.2) -16.6 (0.6) 0.04 (0.02) -1.33 (0.08) Descent 6 -26.7 (1.3) 31.1 (1.0) -30.4 (3.0) 0.03 (0.01) -1.23 (0.04) LW 7 -1.2 (1.5) 16.8 (2.4) -13.6 (1.5) 0.19 (0.03) -1.66 (0.06) DJC Ascent 5 14.1 (1.0) 18.7 (0.9) -21.1 (1.4) 0.02 (0.006) -1.90 (0.69) Descent 4 -16.9 (2.4) 23.3 (0.8) -21.7 (1.0) 0.03 (0.02) -1.26 (0.21) LW 6 -2.3 (2.6) 13.6 (1.4) -12.7 (2.6) 0.20 (0.06) -1.40 (0.10) NJR Ascent 2 12.5 (0.7) 14.5 (0.6) -21.3 (7.0) 0.01 (0.01) -1.48 (0.01) Descent 5 -23.1 (1.2) 29.4 (0.5) -28.2 (1.7) 0.02 (0.00) -2.02 (0.12) LW 5 0.5 (1.7) 13.5 (1.1) -24.3 (4.1) 0.08 (0.02) -1.57(0.03) SEP Ascent 5 6.9 (0.8) 16.8 (0.7) -28.1 (3.2) 0.02 (0.01) -1.35 (0.10) Descent 0 -----LW 7 -2.3 (2.4) 11.7 (1.1) -16.9 (1.4) 0.14 (0.02) -1.50 (0.07) MJK Ascent 2 14.5 (2.2) 23.2 (1.5) -13.2 (0.4) 0.03 (0.02) -1.45 (0.14) Descent 5 -19.9 (1.5) 29.0 (0.8) -24.2 (2.2) 0.02 (0.01) -1.22 (0.12) LW 6 -2.6 (2.8) 9.9 (1.5) -23.8 (6.0) 0.19 (0.03) -1.49 (0.14) AG Ascent 6 14.7 (2.9) 19.6 (3.7) -19.1 (6.1) 0.02 (0.01) -1.44 (0.16) Descent 5 -19.3 (1.5) 35.5 (1.6) -25.9 (3.4) 0.02 (0.00) -1.34 (0.08) LW 8 2.0 (1.8) 17.0 (1.5) -15.1 (2.6) 0.15 (0.02) -1.51 (0.11) KMC Ascent 6 15.7 (1.5) 22.8 (0.9) -13.7 (1.9) 0.02 (0.01) -1.52 (0.06) Descent 5 -22.3 (2.1) 38.9 (2.1) -23.5 (2.7) 0.02 (0.01) -1.39 (0.09)

26

4.1 STAIR ASCENT The kinematic and kinetic patterns at the ankle are significantly different during stair ascent and level walking. Unlike level walking, during stair ascent, the foot always strikes the ground with the forefoot rather than the heel. Because individuals must bend their knee as they lift it, the ankle angle at foot strike is approximately 15o in dorsiflexion (Refer to Table 4.1 and Figure 4.1). During the first portion of stance, the ankle becomes more dorsiflexed as the foot accepts the weight. When the opposite foot comes off the ground, the foot begins to powerfully plantarflex to raise the bodys center of mass and bring the opposite foot up. When the opposite foot hits the stair above it, the second double support phase begins. The center of mass moves further forward, the foot again dorsiflexes slightly and some power is absorbed. During the latter portion of stance, the foot pushes forcefully up off the step and the swing phase begins. Figure 4.2 shows the kinematic and kinetic trajectories for a single trial of a representative subject.

4.2 CHARACTERIZING ANKLE FUNCTION DURING STAIR ASCENT Because the power changes sign three times during stance, stair ascent was separated into four phases. These phases consisted of two controlled dorsiflexion phases where power was absorbed and two powered plantarflexion phases where power was generated. Phases were defined by zero crossings of the ankle angular velocity (from plantarflexion to dorsiflexion and vice versa). Since power is defined as the product of the moment and velocity, these points coincided with the zero crossings of the power. The last phase ended as stance ended, as the foot came off the step. 27

Double support phase Single support phase

Foot off

FF

Figure 4.2 Ankle angle, velocity, moment and power for one cycle of stair ascent. Graphs are shown for a single trial of subject KMC. The circles indicate the zero crossings of the velocity that divide the stance phase into sub-phases.

28

PHASE 1: CONTROLLED DORSIFLEXION 1 The first phase of stair ascent, controlled dorsiflexion 1 (CD1), begins with foot strike. As mentioned above, the foot is in a dorsiflexed position and continues to

dorsiflex until the heel contacts the step, denoted foot flat (FF). During this phase, the power is always negative, thus only passive devices where considered. Scatter plots of ankle torque versus angular velocity did not show a linear relationship between the two variables. The velocity increases during the first portion of the CD1 phase while the moment decreases, then begins to decrease during the latter half of the phase while the moment continues to increase at a constant rate. The angular position, however, continues to increase through the entire phase as the moment decreases. Scatter plots of the ankle torque versus angular position did show a linear relationship between moment and position. Linear regression of the ankle torque on the ankle angular position during this phase was significant for all trials. Spring-like behavior was plausible since a significant amount of power is generated during stair ascent compared to the amount absorbed. It would be far less efficient to use a damper to dissipate this energy since a damper does not provide any energy return. Figure 4.3 shows data from the CD1 phase for the same sample trial of the subject whose data are shown in Figure 4.2 for the entire stair cycle. The ankle torque is a plotted versus the ankle angular position in degrees. This plot shows a strong linear relationship between the moment and position. Average R-squared values for each

subject ranged from 0.84 to 0.98 (See Table 4.2). No clear pattern was seen in the

29

residuals for any trial. During this phase, the ankle behaves as a linear spring with an average stiffness between 0.061 and 0.150 Nm/kg-deg.

Figure 4.3 Results of a linear regression of ankle torque on ankle angle for a single trial of subject KMC during the first controlled dorsiflexion phase of stair ascent (R2=0.98).

30

Table 4.2 Results of a linear regression of ankle torque on ankle angular position during the first controlled dorsiflexion phase of stair ascent. Subject Label SPC ELS HM CLH DJC NJR SEP MJK AG KMC Spring Stiffness # of trials k (Nm/kg-deg) 6 5 6 4 5 2 5 2 6 6 -0.014 (0.093) -0.101 (0.030) -0.150 (0.040) -0.102 (0.047) -0.128 (0.029) -0.087 (0.019) -0.071 (0.018) -0.061 (0.013) -0.127 (0.039) -0.123 (0.024) R
2

# of Points Used for Regression 19 (2.9) 17 (1.3) 19 (5.5) 27 (2.7) 15 (2.4) 10 (1.4) 19 (2.8) 29 (3.5) 18 (4.1) 25 (2.5)

Time in phase (sec) 0.160 (0.025) 0.146 (0.011) 0.162 (0.046) 0.222 (0.023) 0.122 (0.020) 0.073 (0.012) 0.157 (0.023) 0.246 (0.030) 0.154 (0.034) 0.214 (0.021)

0.93 (0.05) 0.96 (0.03) 0.87 (0.09) 0.89 (0.07) 0.91 (0.04) 0.84 (0.08) 0.96 (0.02) 0.91 (0.01) 0.92 (0.04) 0.98 (0.01)

PHASE 2: POWERED PLANTARFLEXION 1 The first powered plantarflexion phase began at foot flat when the ankle reached its maximum dorsiflexion as the angular velocity went from positive to negative. During this phase, the body began its single leg support. To hold the support leg still while swinging the contralateral leg, power had to be generated at the ankle. Since less than half of the work required for the PP1 phase is absorbed during the CD1 phase, it was not possible to use a purely passive device to drive this phase (See Table 4.3). Using the release of the spring from the CD1 phase could move the ankle through a portion of the cycle, but an active torque actuator is needed to assist the spring in moving the ankle the remaining distance. 31

Table 4.3 Summary of work generated during each phase of stair ascent. Work is measured in Joules per kilogram. Subject Label SPC ELS HM CLH DJC NJR SEP MJK AG KMC # of trials 5 5 6 4 4 2 5 2 6 6

WCD1
-1.08 -3.01 -1.16 -0.87 -1.59 -0.13 -3.40 -1.76 -1.44 -3.79

WPP1
7.22 11.20 8.86 4.68 7.35 4.13 6.65 5.39 7.76 8.82

WPP1 |WCD1|
6.14 8.19 7.7 3.81 5.76 4.00 3.25 3.63 6.32 5.03

WCD2
-2.76 -0.75 -5.02 -3.43 -1.98 -1.50 -2.18 -1.84 -2.71 -1.30

WPP2
18.51 20.15 26.95 21.38 14.30 20.77 19.57 20.17 20.02 21.89

WPP2|WCD2|
15.75 19.40 21.93 17.95 12.32 19.27 17.39 18.33 17.31 20.59

Since the ideal solution would be to utilize the elastic return of the spring, the position/torque relationship was explored for the PP1 phase, by fitting the residual torque once the elastic return of the spring had been taken into account. Because it is relatively simple to drive torque actuators by using position data or velocity data, fitting of the residual torque trajectories was accomplished by plotting residuals versus ankle angle and versus ankle angular velocity. The ankle angular position decreased during the entire phase as the moment increased. Scatter plots of the ankle torque versus ankle angular position showed a consistent relationship between torque and angular position for most subjects. This relationship is not strongly linear, however. Average R-squared values ranged from 0.11 to 0.93 (See Table 4.4). For those subjects with low R-squared values, instead of 32

increasing at the beginning the phase, the moment leveled out to a constant value. For these subjects, the beginning of the phase could be modeled using a constant torque damper while the remainder would require a torque actuator. For subjects where the moment increased steadily throughout the phase, the ankle function could be modeled adequately with a linear relationship between torque and ankle angle. Figure 4.4 shows the results of a simple linear regression of ankle torque on ankle angle for the same representative subject.

Ankle Angle (deg) Figure 4.4 Results of a linear regression of ankle torque on ankle angular position during the PP1 phase of stair descent. Data are shown for a single trial of subject KMC (R2=0.95)

33

Table 4.4 Results of a simple linear regression of ankle torque on ankle angular position during the first powered plantarflexion phase of stair ascent Subject Label SPC ELS HM CLH DJC NJR SEP MJK AG KMC # of trials 6 5 6 4 5 2 5 2 6 6 Spring Stiffness k (Nm/kg-deg) -0.007 (0.009) -0.018 (0.002) -0.005 (0.013) -0.010 (0.038) -0.031 (0.009) -0.004 (0.018) -0.040 (0.016) -0.012 (0.025) -0.025 (0.009) -0.039 (0.009) R
2

# of Points Used for Regression 50 (7.5) 46 (10.0) 42 (3.7) 51 (3.6) 45 (2.6) 67 (2.1) 51 (4.1) 51 (7.8) 56 (7.7) 40 (4.7)

Time in phase (sec) 0.418 (0.062) 0.383 (0.084) 0.349 (0.030) 0.425 (0.030) 0.373 (0.022) 0.562 (0.018) 0.423 (0.035) 0.429 (0.065) 0.467 (0.065) 0.331 (0.043)

0.41 (0.27) 0.62 (0.26) 0.48 (0.23) 0.78 (0.11) 0.90 (0.03) 0.11 (0.11) 0.88 (0.07) 0.62 (0.32) 0.66 (0.21) 0.93 (0.03)

PHASE 3: CONTROLLED DORSIFLEXION 2

The third phase of stair ascent begins as the ankle starts to move back into dorsiflexion. During this phase the power is always negative so only passive devices were considered. Again, the possibility of an energy storage release mechanism was explored to fuel the last phase. Scatter plots of ankle torque versus ankle angle showed a linear relationship between the two variables. Figure 4.5 shows this relationship for a single trial of subject KMC. The relationship for all subjects was highly linear with Rsquared values greater than 0.90 for all subjects and residuals evenly distributed around zero (See Table 4.5). The average ankle stiffness for each subject ranged from 0.110 to 0.408 Nm/kg-deg. 34

Table 4.5 Results of a simple linear regression of ankle torque on ankle angle for the CD2 phase of stair ascent. Subject Label SPC ELS HM CLH DJC NJR SEP MJK AG KMC # of trials 6 5 6 5 5 2 5 2 6 6 Ankle Stiffness k (Nm/kg-deg) -0.161 (0.056) -0.410 (0.121) -0.163 (0.062) -0.110 (0.027) -0.292 (0.097) -0.353 (0.059) -0.330 (0.371) -0.175 (0.025) -0.179 (0.123) -0.408 (0.186) R2 0.95 (0.04) 0.93 (0.07) 0.92 (0.06) 0.96 (0.02) 0.97 (0.01) 0.97 (0.02) 0.94 (0.05) 0.95 (0.02) 0.93 (0.03) 0.97 (0.01) # of Points Used for Regression 24 (4.3) 10 (4.0) 21 (4.6) 29 (3.4) 16 (1.6) 21 (0.0) 22 (5.6) 18 (0.7) 23 (7.2) 16 (2.4) Duration of Phase (sec) 0.202 (0.036) 0.085 (0.034) 0.173 (0.038) 0.243 (0.028) 0.132 (0.014) 0.175 (0.000) 0.185 (0.050) 0.154 (0.006) 0.192 (0.060) 0.133 (0.020)

35

Figure 4.5 Scatter plots of ankle torque versus ankle angle for the CD2 phase of stair ascent. Data are shown for the same trial as Figure 4.4, R2= 0.97.

36

PHASE 4: POWERED PLANTARFLEXION 2 During the second powered plantarflexion phase, the foot pushes off the step. This phase ends as the foot comes off the ground and stance ends. During stair ascent, the amount of energy required to push the foot off the ground is less than in level walking. The reason for this, is that the center of pressure at foot off is close to the ankle since the stride length during stair walking is decreased. The amount of energy absorbed during the CD2 phase was, on average, only one tenth of the amount generated during the PP2 phase. As the power absorbed was

insufficient to fuel this movement, only active devices were considered. Scatter plots of ankle torque verses ankle angular position showed a strong linear relationship between the two variables with the exception of the first 8-10 points (Figure 4.6). This portion of the phase can be modeled using the return of the energy stored in the spring during the CD2 phase. The remainder of the phase requires energy input from an active device. It can therefore be modeled using a torque actuator that increases the amount of torque produced as ankle angle increases.

37

CD2 phase

Return from spring

Figure 4.6 Plots of ankle torque on ankle position for the CD2 and PP2 phase of stair ascent. Inset shows portion of phase where energy is stored and released in the spring. Data are shown for a single trial of subject KMC.

38

Table 4.6 Results of a linear regression of ankle torque on ankle angle for the second powered plantarflexion phase of stair ascent. Subject Label SPC ELS HM CLH DJC NJR SEP MJK AG KMC # of trials 6 5 6 4 5 2 5 2 6 6 Ankle Stiffness k (Nm/kg-deg) -0.049 (0.014) -0.047 (0.007) -0.064 (0.013) -0.043 (0.002) -0.077 (0.020) -0.055 (0.013) -0.047 (0.006) -0.052 (0.006) -0.052 (0.009) -0.053 (0.004) R
2

# of Points Used for Regression 22 (3.4) 22 (2.9) 20 (2.1) 26 (1.7) 17 (2.2) 25 (5.7) 23 (2.3) 23 (0.0) 21 (2.5) 28 (1.6)

Time in phase (sec) 0.183 (0.028) 0.181 (0.032) 0.164 (0.017) 0.215 (0.014) 0.140 (0.018) 0.208 (0.047) 0.193 (0.019) 0.192 (0.000) 0.175 (0.020) 0.232 (0.013)

0.91 (0.02) 0.88 (0.05) 0.92 (0.04) 0.93 (0.03) 0.91 (0.07) 0.92 (0.01) 0.93 (0.05) 0.79 (0.01) 0.88 (0.09) 0.94 (0.03)

4.3 MODEL OF ANKLE DURING STAIR ASCENT In summary, during stair ascent, the ankle functions as a system of springs and torque actuators. Figure 4.7 shows a schematic of this model. During the first portion of stance, CD1, the ankle behaves as a linear spring. During the second portion, powered plantarflexion 1 (PP1), a torque actuator assists the energy returned from the spring in dorsiflexing the foot in seven of the ten subjects. In the remaining three subjects, the ankle is modeled using a constant torque damper, which requires active control to hold the torque at a constant level. In the next phase, controlled dorsiflexion 2 (CD2), the ankle is again modeled using a linear spring. In the final phase of stance, the ankle was modeled as a torque actuator that assists the spring in plantarflexing the foot. Finally, the

39

foot comes off the step and the body enters the swing phase (SW) where the same model used for level walking could be fine-tuned for stair ambulation.

T OR

CD

PP1
b

SW

PP2
T

CD2

Figure 4.7 Schematic of the model of ankle function during stair ascent. 4.4 STAIR DESCENT The requirements of descending stairs are significantly different than that of level walking and stair ascent. When descending stairs, the foot strikes the step in a more plantarflexed position where the center of pressure is on the forefoot rather than the heel (Refer to Table 4.1). During initial stance the foot begins to accept the weight of the body. As the body moves from a higher position, a significant amount of potential energy is absorbed. This energy causes the foot to dorsiflex passively as the weight is shifted from the ipsilateral leg. At the time of opposite foot contact, the foot is at its 40

maximum dorsiflexion position.

At this time, the foot requires power to forcibly

plantarflex the foot and push off the step. Figure 4.8 shows the ankle angle, velocity, moment and power for one cycle of stair descent. CD1

CD2

PP

Figure 4.8 Position, velocity, moment and power at the ankle for a single gait cycle of stair descent. Data are shown for a single trial of subject MJK.

41

4.5 CHARACTERIZING ANKLE FUNCTION DURING STAIR DESCENT The stance phase of descent was divided into three sub-phases. The first phase began at foot strike and lasted until the heel made contact with the step (foot flat). During this time phase, the foot accepts the weight of the body and power is absorbed. Since the foot moves into dorsiflexion as the heel come down, this phase was called controlled dorsiflexion (CD1). At foot flat, the ankle began the weight transfer portion of stance where the body weight is transferred to the leg as the contralateral foot comes off the ground. During this period, the shank comes forward so the ankle angle continues to increase in degree of dorsiflexion. Since power was also absorbed in this phase, it was called the second controlled dorsiflexion phase (CD2). When the ankle reached its maximum dorsiflexion position, the foot began the only powered phase of stance. This sub-phase, referred to as powered plantarflexion (PP) lasted until the foot came off the step (FO). PHASE 1: CONTROLLED DORSIFLEXION 1 The first controlled dorsiflexion phase begins with foot strike. At this point the foot is in a plantarflexed position and the external moment and power are about zero. During this phase, the foot moves into dorsiflexion while there is a plantarflexor moment. Since the angular movement and torque are in opposite directions, power is absorbed during this phase. As such only passive devices (dampers and springs) were considered. Since the energy absorbed in the second phase (ie. controlled dorsiflexion 2) was greater than the amount generated in the third phase (powered plantarflexion) by 2.3 to 11.2 J/kg, it was most efficient to view these phases as the energy storage and release of a 42

spring with additional damping. In view of the fact that a spring absorbs energy that needs to be released at some later point in the cycle, a spring mechanism which is advantageous for modeling control dorsiflexion 2 would be problematic for modeling control dorsiflexion 1 and a damping device appears to be preferable. This phase could not adequately be modeled using a simple linear damper, however. Plots of ankle velocity versus ankle torque showed that ankle velocity

increased with ankle torque to a point, then began to decrease. Since a simple linear damper was insufficient, magnetorheological dampers were considered. These dampers can vary their behavoir in any desired pattern by altering the magnetic field applied to the magnetorheological fluid contained in the device. To control the pattern of the dampening, a relationship (linear or not) needs to be found between ankle torque and ankle angle, velocity, or both variables. Plots of ankle torque versus ankle angle and ankle torque versus ankle velocity showed consistent relationships among trials and among subjects. Figure 4.9 shows plots of ankle torque versus ankle angle and ankle torque versus ankle velocity for two subjects. From a linear regression of ankle torque on ankle angle, average R-squared values for each subject were computed. These values varied between 0.30 and 0.96 depending on the duration of the time-interval during which the ankle torque became constant at the end of the CD1 phase. In order to model this phase, equations would need to be found that related changes torque to changes in position and velocity. These relationships could be determined on an

individual basis as there is less variability within subjects than between them (Graphs for all subjects are given in Appendix A). 43

Figure 4.9 Plots of ankle torque versus ankle angle and velocity for the CD1 phase of stair descent. Data are shown for subjects KMC and NJR.

PHASE 2: CONTROLLED DORSIFLEXION 2 The second sub-phase of descent began at foot flat where the ankle position reached a plateau. Since this is an inflection point rather than a local minimum or maximum, it was located by the zero crossing of the acceleration. For a few subjects, a small amount of power was generated during this sub-phase. For those subjects, the ankle began to plantarflex slightly, and then return to dorsiflexion. This occurred on a few of the subjects for a single trial but only consistently in subject ELS. For a majority of the subjects the plots of ankle angle versus time decreased in slope rather than 44

completely changing directions. Although the power went slightly positive in a few of these cases, the power for a majority of the phase was negative. As mentioned above, the amount of power absorbed during the CD2 phase was compared to the amount generated during the PP phase to determine if the energy stored was sufficient to drive the system. Work Required = WPP |WCD2| (4.1)

Results of this analysis are given in Table 4.7. Since the required work was always negative, it would be most efficient to model the ankle as a system that could store and release energy. With this idea, the ankle torque / angle relationship was explored to determine if the ankle could be modeled as a spring during this phase.

Table 4.7 Work done during each phase of stair descent1 Average Average Average Subject # of WCD2 W CD1 WPP Label trials (J / kg) (J / kg) (J / kg) SPC 3 -17.89 -13.13 9.42 ELS 4 -23.31 -9.33 10.44 HM 5 -25.99 -14.79 11.32 CLH 6 -17.91 -19.11 7.92 DJC 4 -24.50 -8.35 6.03 NJR 4 -59.67 -10.96 7.80 MJK 5 -21.59 -12.75 10.84 AG 5 -27.83 -18.46 10.44 KMC 5 -31.97 -12.09 9.50

WPP -|WCD2| (J / kg) -3.72 -3.72 -3.47 -11.20 -2.33 -3.16 -1.90 -8.02 -2.58

Subject SEP was eliminated from analysis of stair descent due to excessive noise in the data that prevented the identification of the beginning and end of cycles and consistent results for parameter values.

45

Ankle torque decreased at a constant rate as the ankle angle increased. Scatter plots of ankle torque versus ankle angle revealed a significant correlation between the two variables. Results of a linear regression with residuals are shown for a single trial of subject CLH in Figure 4.10. Analysis of the residuals did not support the assumption of a linear spring however. The residuals were not randomly dispersed around zero. As it would be impractical to show residual plots for each trial, results of a residual analysis are included in Table 4.8. The fit was evaluated by looking at the number of times the residual changed sign versus the number of points used in the regression. If the pattern were random, one would expect this ratio to be around 0.5. For a linear regression of ankle torque on ankle angular position, the ratios were between 0.05 and 0.27 indicating that a linear fit is insufficient to explain the ankle behavior during this phase.

46

Figure 4.10 Results of a simple linear regression of ankle torque on ankle angular position during the second controlled dorsiflexion phase of stair descent including plots of the residuals. Results are shown for a single trial of subject CLH (R2=0.93).

47

Table 4.8. Results for a simple linear regression of ankle angle on ankle torque for the second controlled dorsiflexion phase and powered plantarflexion phase of stair descent Subject Phase SPC CD2 PP CD2 PP CD2 PP CD2 PP CD2 PP CD4 PP CD2 PP AG CD2 PP CD2 PP 5 Ankle # of Stiffness, k trials (Nm/kg-deg) 5 R2 # of points used for regression 49 (2.5) 22 (3.6) 30 (6.5) 17 (0.5) 52 (12) 15 (2.2) 67 (7.8) 25 (0.8) 39 (1.9) 15 (2.7) 52 (3.5) 16 (1.0) 43 (3.2) 21 (1.8) 46 (6.0) 17 (0.7) 42 (4.7) 17 (1.2) Time (sec) Residuala

-0.036 (0.00) 0.91 (0.02) -0.060 (0.01) 0.96 (0.02) -0.035 (0.01) 0.68 (0.30) -0.067 (0.02) 0.99 (0.00) -0.023 (0.01) 0.45 (0.30) 0.080 (0.01) 0.99 (0.00) -0.031 (0.01) 0.86 (0.09) -0.041 (0.01) 0.92 (0.02) -0.065 (0.01) 0.88 (0.06) -0.093 (0.01) 0.99 (0.02) -0.036 (0.01) 0.80 (0.10) -0.083 (0.02) 0.99 (0.00) -0.037 (0.01) 0.87 (0.11) -0.050 (0.01) 0.94 (0.02) -0.032 (0.01) 0.92 (0.05) -0.057 (0.00) 0.96 (0.02) -0.030 (0.01) 0.93 (0.06) -0.058 (0.01) 0.97 (0.01)

0.404 (0.02) 0.27 (0.06) 0.183 (0.03) 0.16 (0.13) 0.28 (0.07) 0.06 (0.02)

ELS

0.140 (0.00) 0.19 (0.07) 0.431 (0.10) 0.07 (0.06) 0.125 (0.02) 0.21 (0.11) 0.561 (0.07) 0.09 (0.04) 0.207 (0.01) 0.09 (0.03) 0.327 (0.02) 0.07 (0.01) 0.123 (0.02) 0.20 (0.11) 0.430 (0.03) 0.06 (0.00) 0.133 (0.01) 0.17 (0.11) 0.363 (0.03) 0.05 (0.01) 0.178 (0.02) 0.09 (0.01) 0.412 (0.04) 0.07 (0.03) 0.142 (0.01) 0.14 (0.05) 0.350 (0.04) 0.15 (0.08) 0.140 (0.01) 0.16 (0.07)

HM

CLH

DJC

NJR

MJK

KMC
a

The residual value is the ratio of the number of changes in sign of the residual over the number of points used in the regression

48

PHASE 3: POWERED PLANTARFLEXION The last phase of stance for stair descent was powered plantarflexion. As

mentioned above, this phase began at the maximum position of the ankle and ended at foot off (FO). To utilize the energy stored in the CD2 phase, the PP phase was modeled as a spring. Figure 4.11 shows the scatter plot of ankle torque versus ankle angle during the PP phase. Examination of the residuals showed a non-random pattern. In Figure 4.12, the residual value is plotted versus the angular velocity and a strong linear correlation is seen. This indicates that during the PP phase, the ankle acts as a

combination of a damper and a spring. Using multiple linear regression with the least squares method, a model was found that adequately represented the ankles behavior during this phase. The ankle behaves as a spring and a damper in parallel given by the following equation,

= k + b

d + o dt

(4.1)

This model reduces the error in the approximation of torque trajectories to under 0.1 Nm/kg. Results of this analysis are shown in Figure 4.13.

49

Figure 4.11 Results of a linear regression of ankle torque on ankle angle during the powered plantarflexion phase of stair descent. Graphs are shown for the same trial as figure 4.10 (R2=0.94)

50

Figure 4.12 Results of a linear regression of the residuals of the ankle angle and ankle torque regression on ankle angular velocity for powered plantarflexion. Data are shown for the same trial shown in Figure 4.11.

51

Figure 4.13 Scatter plots of ankle torque versus ankle angle for a spring and damper model of the powered plantarflexion phase of stair descent. Results are shown for the same subject. While these models are the best fit for each phase individually, the stiffness required in each phase is not the same. In order to get an energy return, the same spring must be used for both phases. Since the stiffness required in the second phase (ie. powered plantarflexion) is greater than that of the first (controlled dorsiflexion 2), the entire system could be modeled using a linear spring with some dampening. Figure 4.14 shows plots of the ankle torque/ ankle angle relationship for the CD2 phase and the residual when a spring with stiffness determined based on optimal fitting of the PP phase is utilized. The residual is plotted as a function of ankle angle as well as ankle angular velocity. The residual does not demonstrate a linear relationship with the 52

ankle position or the angular velocity. In order to design a device that functioned in this way, the same variable damper employed in the CD1 phase would have to be used. This damper would change dampening during the CD2 phase to compensate for the nonlinearity and remain constant for the PP phase as shown above. More quantitative analysis is necessary to find an equation (between ankle torque and ankle position, velocity or both variables) to drive the variable damper during the CD2 phase.

53

Figure 4.14 Scatter plot of ankle angle verse ankle torque for the CD2 phase of stair descent. Residuals values for a fitting using the spring stiffness defined by the PP phase are shown versus ankle angle and velocity. Data are shown for a single trial of subject MJK.

54

4.6 MODEL OF THE ANKLE DURING STAIR DESCENT In summary, the ankle was modeled using a variable damper and linear spring during stair descent. During the first portion of the stance phase, controlled dorsiflexion 1 (CD1) the ankle behaves as a variable damper as it dissipates the potential energy absorbed when stepping down. In the subsequent two phases, controlled dorsiflexion 2 (CD2) and powered plantarflexion (PP), the ankle is modeled with a variable damper and linear spring in parallel. The damper varies its mechanical characteristics during the CD2 phase and maintains a constant damping coefficient during the PP phase. Finally, as the foot comes off the step, the foot enters the swing phase (SW).

CD1

SW CD2 CD2

PP

Figure 4.15 Schematic diagram of the model of ankle behavior during stair descent.

55

CHAPTER 5. MODELING TRANSITIONS In addition to modeling the stance phase of level walking, stair ascent and descent, it is necessary to look at the transitions between these tasks in order to design a prosthesis or orthosis adequately. The ideal design allows for smooth transitions not only during different tasks, but also during the period when movement changes from one pattern to the next. To this end, the transition from descent to level walking and the transition from level walking to stair ascent were analyzed. As there was no

instrumented platform at the top of the stairway, the transition from ascent to level walking and level walking to descent were not studied.

5.1 ASCENT TO LEVEL WALKING The transition from level walking to stair ascent was observed during the stance phase of the foot contacting the plate (just prior to the steps). This pattern differed from level walking in that there was little initial plantarflexion of the foot. In two of the ten subjects, there was no initial plantarflexion. In those subjects where plantarflexion was seen, the moment at the ankle was approximately zero. For this reason, it was not possible to distinguish a controlled plantarflexion phase like that seen in level walking. One reason for this difference might be the way the subjects initiated the movement. Rather than walking up to the stairs and ascending them, the subjects were asked to take a single step onto the force platform and walk up the steps. While this is only one of the possible ways to approach a flight of stairs among all natural patterns of transition

56

between level walking and stair ambulation, it was the only way to measure the forces given the limitations of the lab set-up.

Figure 5.1 Ankle angle, velocity, torque and power for the transition from level walking to stair ascent. Data are shown for one trial for subject SEP. 57

Since the subjects were only taking one step on level ground, they began closer to the platform, which shortened the stride length. For this reason, the bodys center of mass at foot strike was almost directly over the foot rather than behind it. As such, the moment did not move into dorsiflexion, but remained at a constant zero level for initial stance. Since, no controlled plantarflexion was seen, the transition from level walking to stair ascent was separated into two phases, controlled dorsiflexion and powered plantarflexion.

CHARACTERIZING THE TRANSITION FROM LEVEL WALKING TO STAIR ASCENT During the controlled dorsiflexion phase, the moment was always negative (plantarflexion) and the angular position was always positive (dorsiflexion). As a result, the power was negative for this entire period of time. Unlike level walking, the moment exhibited a biphasic pattern. Rather than decreasing steadily until opposite foot contact, the moment decreased during initial stance, increased and then decreased again before opposite foot contact. The power absorbed during this phase was significantly less than in level walking. Since the power is always negative, only springs and dampers were considered for modeling this transition phase. Since power generation is required in the subsequent, push-off, phase, spring-like behavior was expected. Scatter plots of ankle torque versus ankle velocity confirmed this hypothesis showing no correlation between the two variables (Figure 5.2).

58

Ankle Velocity (deg/s)

Figure 5.2 Scatter plot of ankle torque versus ankle velocity for the transition from level walking to stair ascent. Data are shown for the same trial of subject SEP. Strong correlations between ankle torque and ankle angular position were seen for seven of the ten subjects (R2>0.8). One subject, ELS, was not included in this analysis because there were no clean hits of the force plate for this subject (part of the weight was on the floor). This was an issue in reducing the number of good trials since the area of the force platform below the stairs was so small. The relationship between ankle torque and ankle angle was not linear as the residuals were not evenly distributed about zero. In addition, there was no correlation between the residual value and the ankle velocity. The difference between the model and the ankle behavior could therefore not be accounted for simply by adding a linear damper.

59

Figure 5.3 Results of a linear regression of ankle torque on ankle angle for the transition from level walking to stair ascent. Data are shown for the same trial as Figures 5.1 and 5.2 (R2=0.92) To account for the difference between the model and the ankle behavior, a variable damper could be employed. The ankle behavior is more consistent between different trials of the same subject than between subjects. It is therefore recommended that an equation for the variable damper is found an individual basis (See Appendix B.) In order to model the specific transition from level walking to stair ascent, the transition must be studied in a more realistic fashion. With the current laboratory set-up, the force platform is directly below the stairs and extends only 13.5 inches. If the lab was 60

equipped with multiple force platforms for example, the subject could walk up and ascend the steps without having to concentrate on foot placement. This would most likely show a pattern in the first sub-phase that is more consistent with level walking.

Table 5.1 Parameter values for a linear regression of ankle torque on ankle angle for the transition from level walking to stair ascent Subject Label SPC HM CLH DJC NJR SEP MJK AG KMC # of trials 4 5 5 3 4 5 4 5 4 Ankle Stiffness k Nm/kg-deg -0.04 (0.02) -0.07 (0.02) -0.04 (0.00) -0.05 (0.02) -0.08 (0.01) -0.06 (0.02) -0.07 (0.00) -0.06 (0.01) -0.05 (0.00) R2 0.71 (0.16) 0.82 (0.06) 0.81 (0.13) 0.54 (0.15) 0.85 (0.09) 0.92 (0.07) 0.85 (0.04) 0.93 (0.03) 0.85 (0.04) # of points used for regression 87 (7.4) 86 (8.0) 108 (5.4) 74 (6.0) 84 (3.3) 91 (8.3) 89 (6.3) 86 (12.4) 86 (3.6) Time (sec) 0.725 (0.062) 0.712 (0.067) 0.898 (0.045) 0.613 (0.050) 0.698 (0.028) 0.755 (0.069) 0.740 (0.052) 0.715 (0.103) 0.717 (0.030)

CHARACTERIZING THE TRANSITION FROM LEVEL WALKING TO STAIR ASCENT During the final phase of the transition, power was generated at the ankle to push off from the ground up to the second step. In this phase the ankle behaves in the same way it does during stair ascent. Since so little power is absorbed at the ankle in the preceding phases, a majority (>60%) of the energy to supply this phase must be generated from a torque actuator. Throughout powered plantarflexion, ankle torque is directly proportional to the ankle position, so position could be used to drive the actuator.

61

Figure 5.4 shows a schematic diagram of stair ascent and level walking. Once the ankle-foot complex moves from the first phase of level walking, CP, it can either continue in the level walking cycle or move into a transition phase. This transition phase is modeled using a variable damper as described above. Once it moves through this phase, the ankle-foot complex moves into the third phase of stair ascent (CD2). From this point the ankle-foot complex continues around the stair ascent cycle until another transition is made.

62

CD1 CP

SW

Level Walking

CD CD2

Transition

PP
T

CD1

PP1

T OR

SW

Stair Ascent

PP2
T

CD2

Figure 5.4 Schematic of the models of ankle function during stair ascent, level walking and the transition from level walking to stair ascent.

63

5.2 DESCENT TO LEVEL WALKING The transition from descent to level walking was observed as the foot coming from the second step struck the force platform on the ground just in front of the steps. (The contralateral leg was on the first step and showed its transition during the swing phase of gait.) As the foot hit the plate it began the cycle in a typical descent pattern with the foot in plantarflexion and power being absorbed. About 10% into the gait cycle, when the opposite foot came off the step above it, the angle and moment showed a slight plantarflexion pattern while little power was absorbed. Since the ankle did not continue to dorsiflex, the peak angle during the transition was less than during stair descent. When the contralateral leg struck the floor, the foot began to plantarflex and the angular movement, moment, and power all began to behave like level walking. Since power changed signs three times, the transition was split into three subphases. The first sub-phase, controlled dorsiflexion, was identical to the CD1 phase of stair descent. This phase began with foot strike and ended when the ankle reached its maximum dorsiflexed position. During the second sub-phase, mid-stance, the transition occurred. As the foot was moving into plantarflexion and power was absorbed, this phase was called controlled plantarflexion. The final phase was powered plantarflexion. During this phase, the ankle generated power to push off the ground. The ankle behavior during powered plantarflexion was identical to that of the powered plantarflexion phase of level walking (Refer to Figure 5.6).

64

Figure 5.5 Graphs of ankle angle, velocity, torque, and power for the transition from stair descent to level walking. Data are shown for a single trial of stair descent and one trial of level walking for subject DJC.

65

CHARACTERIZING THE TRANSITION FROM DESCENT TO LEVEL WALKING As mentioned above, the controlled plantarflexion, or transition phase, began when the ankle reached its maximum position (when the velocity crossed zero) and ended when the ankle velocity crossed from positive to negative a second time. During this

period the ankle power is predominately positive. Since the first phase continues from stair descent, it is modeled using a variable damper. As such there is no energy storage to supply this phase. The transition must therefore be fueled using a torque actuator. The behavior of the actuator during this period is non-linearly related to ankle position or velocity. So a combination of the two variables is necessary. As seen in the transition from level walking to stair ascent, there is a great deal of variability in the data due to the numerous ways the transition can occur. Unlike stair ascent however, this variability also occurs between different trials of the same subject. Appendix C shows plots for each subject during this phase. Figure 5.6 shows a schematic diagram of the models for stair descent and level walking. The ankle-foot complex begin functioning in a stair descent cycle. Once the ankle moves through the first phase of stair descent (CD1), it can either continue into the next phase of descent (CD2) or move into a transition phase. The transition phase is driven actively by a torque actuator. Once the ankle-foot complex moves through the transition phase it enters the level walking cycle at the push-off phase (PP). It then continues in the level walking cycle until another transition is made.

66

Figure 5.6 Schematic of the ankle model during stair descent, level walking and the transition from stair descent to level walking

CP CD1

SW

Level Walking

CD CD2

PP
T

CD1

Transition

SW

Stair Descent

CD2 CD2

PP2

67

Chapter 6. Discussion In this project it was demonstrated that stair walking can be modeled using simple mechanical elements and that stair cycles can be separated into phases using information from simple devices (i.e. sensing ankle angle, velocity or acceleration). Kinematic and kinetic data collected for these tasks showed that it is possible to locate transitions among stair cycle sub-phases based on sensing capability as that provided by goniometers and accelerometers. It was also shown that each phase of stair ambulation can be modeled using simple mechanical elements such as linear springs and dampers or torque actuators driven by sensing ankle angle or angular velocity. Where these linear elements were insufficient, the ankle was modeled using slightly more complex, variable dampers and torque actuators.

6.1 LOCATING TRANSITIONS In order to utilize the model of ankle function for different tasks, there must be a way to switch from one task to another. In this research, we collected kinematic data (position, velocity), accelerometer data and electromyography data to attempt to locate this transition point. Since the transitions could adequately be detected using purely kinematic and kinetic data, the accelerometer and electromyographic (EMG) data were not examined. This is a preferable method since position and velocity sensors are

relatively small, inexpensive devices that could be part of the device itself. EMG data could be examined to see how the muscles behave during periods were the body transitions from one movement to another. Using surface EMG electrodes to 68

detect transitions in movement, however, is a rather obtrusive approach to monitoring transitions. Surface EMG electrodes need to be in direct contact with the skin in order to measure muscle activity. If they are displaced even slightly on the muscle body, the timing and amplitude of the measured signal will change. Also, the skin must be clean and dry in order to reduce noise in the signal. Another possibility would be the use of fine wire electrodes. These electrodes are inserted into the muscles belly with a needle. The problem with these electrodes is that they are uncomfortable for the subject. As an alternative to kinematics and kinetics of the ankle, accelerometers could be used to detect transitions in movement. Accelerometer data were collected via four threeaxis accelerometers. The accelerometers were placed 10 cm above the ankle and knee joints bilaterally. With accelerometer data, we might be able to detect the transitions before the ankle behavior changes, giving the device time to change state. For instance, the accelerometers could detect the movement of the knee upward indicating that the ankle device should switch to a stair ascent pattern. Although there was no time in this thesis to explore the accelerometer data, it is something that should be explored in future works.

6.2 VALIDITY OF THE MODEL It is important to note that although the fitting of ankle torque to ankle angle and velocity data are significant with R-squared values greater than 0.80 and in most cases greater than 0.95, there is a possibility that the accuracy of the approximation obtained by these models may not be sufficient to allow for biologically realistic ambulation. There 69

are two ways to test whether the models developed in this thesis are adequate to design prosthetic and orthotic systems and provide the expected improvement upon currently available systems. A first approach is to develop a forward dynamics model to test the effect of discrepancies between approximated kinematic and kinetic trajectories obtained using the developed models and actual trajectories as measured from trials recorded from individuals. The analyses would include studies of stability of energy consumption as they would be affected by changes in kinematic and kinetic trajectories due to the discrepancies between target values (i.e. natural trajectories) and values obtained by approximating them with the developed models. The second possible approach is to build devices and test their behavior on subjects. The latter is the approach utilized so far by Dr. Herrs team. They first developed models of ankle biomechanics with Palmers work and then applied them to the design of an orthotic device implemented to allow level walking at different speeds. Our plan is to proceed in a similar way but extend the investigation to stair ambulation. Therefore, the mechanical elements and controller of the existing ankle-foot orthosis designed by Dr Herrs team will be modified to allow testing for stair ambulation. By using an experimental approach to testing the adequacy of the developed models as described above, a number of other potential limitations will be tested as well. For instance, we will be able to assess how errors associated with the estimation of ankle spring stiffness impact gait patterns. The model adopted for the analysis performed in this thesis makes several assumptions upon which the stiffness is derived. First, it is assumed that the system behaves as a rigid body, when in fact, there is bending in the 70

foot and changes in the foot shape during movement. Second, the ankle joint center is also assumed to be constant throughout all movements when in reality its position varies over time. This movement was assumed to have a negligible effect on the ankle torque versus ankle angle and angular velocity characteristics since it is small relative to the moment arm (distance from the center of pressure to the ankle joint center). As

mentioned previously, Palmers model for level walking was utilized to design ankle-foot orthosis that functioned in a biologically realistic way [20, 23]. Since the same methods were used to develop these models for stair walking, it is expected that the spring stiffness estimates are sufficiently accurate to design an orthotic or prosthetic device for stair ambulation. In addition to the models developed for stair ambulation, the transitions between level walking and stair ascent and stair descent and level walking were located and modeled. The robustness of the solution proposed in this thesis could be improved, however. An important point to mention is that these are only possible ways in which a person can make transitions between tasks. Subjects were asked to take a single step onto the force platform and then climb the stairs for stair ascent. They were positioned directly in front of the stairs and began movement from stance. In practical situations, a person can walk directly to a staircase and then ascend it; they could turn the corner around a multiple flight staircase and then walk up, or begin from standing. The same is true of descent. For this task, subjects were asked to descend the stairs, step on the plate, and then continue walking across the walkway. However, for most subjects, it would

71

probably have been more natural to step further away from the bottom step so as not to reduce their stride length. To make the design more robust, various possible transitions would need to be examined to determine how they differ. Each variation could change the model in some way. As the laboratory at Spaulding Rehabilitation Hospital is not currently set up to measure forces anywhere except directly below the stair case, further examination was not possible in this work.

6.3 SIGNIFICANCE OF THE WORK This research provides important information to the design of ankle-foot orthosis and prosthesis. First, the models derived in this thesis require mechanical elements that are not needed in level walking, indicating that dramatic changes in orthotic and prosthetic design are needed to enhance individuals mobility beyond level walking. Limitations of currently available technology are emphasized by the results of this research. For instance, prosthetic ankle-foot systems made from carbon fiber springs are not optimal for tasks such as stair descent, where damping is necessary. As seen in Palmers study of walking at different gait speeds, it was shown that active control is necessary in stair ascent [20]. Therefore devices designed using purely passive systems are insufficient to replicate the ankle behavior during stair ascent. The models developed in this work could be used to supplement those models developed previously for level walking. Using a state machine as a controller, such as that seen in Blaya and Herrs active ankle-foot orthosis, existing prototypes of adaptive 72

orthoses can be modified to test the models derived in this project for stair ascent and descent. In fact, existing state machine controllers can be modified by adding states according to the sub-phases of stair ambulation defined in this thesis. While

modifications to the parameter values may be necessary, the specific devices needed to mimic normal human movement have been identified through this research.

6.4 SUGGESTIONS FOR FUTURE WORK An optimal ankle prosthetic or orthotic device needs to account for any activity in which the person is engaged. The next step in this research would be to model the ankle behavior in the same way for a variety of tasks such as walking on ramps or through uneven terrain. In addition, the transitions between all of these movements must be examined in a variety of ways, as a non-disable individual is capable of transitioning from one task to another quickly and in countless possible ways. With the models developed in this project, future ankle prosthetic and orthotic systems could allow more biologically realistic patterns of movement and thus enhance mobility in individuals.

73

Appendix A: Graphs of ankle torque versus ankle angle and velocity for each subject during the CD1 phase of stair descent

74

75

Appendix B. Graphs of ankle torque versus ankle angle and velocity for the transition from level walking to stair ascent

SPC

HM

CLH

DJC

NJR

76

KMC

MJK

SEP

AG

77

Appendix C. Graphs of ankle torque versus ankle angle and velocity for the transition from stair descent to level walking

78

79

WORKS CITED 1. Andriacchi, T.P., J.O. Galante, and R.W. Fermier, The influence of total kneereplacement design on walking and stair-climbing. The Journal of Bone and Joint Surgery American Edition, 1982. 64(9): p. 1328-35. Kay, H.W. and J.D. Newman, Relative incidences of new amputations. Orthothotics and Prosthetics, 1975. 29(2): p. 3-16. Muilenburg AL, Wilson, AB. A Manual for Below-Knee (Trans-Tibial) Amputees. Linthicum, MD: Dankmeyer, Inc.; 1996. Available at: http://www.oandp.com/resources/patientinfo/manuals /bkindex.htm. Accessed October 23, 2003. Brown, S.J., Marketing trends in managed care. 1995, Baylor Institute for Rehabilitation: Dallas, TX. Teasell, R.W., et al., Physical and functional correlations of ankle-foot orthosis use in the rehabilitation of stroke patients. Archives of Physical Medicine & Rehabilitation, 2001. 82(8): p. 1047-9. Hesse, S., et al., Non-velocity-related effects of a rigid double-stopped ankle-foot orthosis on gait and lower limb muscle activity of hemiparetic subjects with an equinovarus deformity. Stroke, 1999. 30(9): p. 1855-61. Rao, S.S., et al., Segment velocities in normal and transtibial amputees: prosthetic design implications. IEEE Transactions on Rehabilitation Engineering, 1998. 6(2): p. 219-26. Klute, G.K., C.F. Kallfelz, and J.M. Czerniecki, Mechanical properties of prosthetic limbs: adapting to the patient. Journal of Rehabilitation Research & Development, 2001. 38(3): p. 299-307. Smith, D.G. and E.M. Burgess, The use of CAD/CAM technology in prosthetics and orthotics-current clinical models and a view to the future. Journal of Rehabilitation Research & Development, 2001. 38(3): p. 327-34. Siegler, S., R. Seliktar, and W. Hyman, Simulation of human gait with the aid of a simple mechanical model. Journal of Biomechanics, 1982. 15(6): p. 415-25. Pandy, M.G. and N. Berme, Synthesis of human walking: a planar model for single support. Journal of Biomechanics, 1988. 21(12): p. 1053-60. 80

2. 3.

4. 5.

6.

7.

8.

9.

10. 11.

12. 13. 14. 15. 16. 17. 18. 19.

Gilchrist, L.A. and D.A. Winter, A two-part, viscoelastic foot model for use in gait simulations. Journal of Biomechanics, 1996. 29(6): p. 795-8. Blickhan, R., The spring-mass model for running and hopping. Journal of Biomechanics, 1989. 22(11-12): p. 1217-27. McMahon, T.A. and G.C. Cheng, The mechanics of running: how does stiffness couple with speed? Journal of Biomechanics, 1990. 23 Suppl 1: p. 65-78. Farley, C.T., J. Glasheen, and T.A. McMahon, Running springs: speed and animal size. Journal of Experimental Biology, 1993. 185: p. 71-86. Farley, C.T. and D.C. Morgenroth, Leg stiffness primarily depends on ankle stiffness during human hopping. Journal of Biomechanics, 1999. 32(3): p. 267-73. Herr, H. and T.A. McMahon, A trotting horse model. International Journal of Robotics Research, 2000. 19: p. 566-581. Seyfarth, A., Geyer, H. Gunther, M. and Blickman, R. A movement criterion for running. Journal of Biomechanics, 2002. 35: p. 649-655. Hansen, A., Childress,D. Miff, S. Gard, S. and Mesplay, K., The human ankle during walking: implications for the design of biomimetric ankle prostheses. Journal of Biomechanics, 2004. In Press. Palmer, M., Sagittal plane characterization of normal human ankle function across a range of walking gait speeds, Masters Thesis, Massachusetts Institute of Technology, 2002. Dingwell, J.B., B.L. Davis, and D.M. Frazier, Use of an instrumented treadmill for real-time gait symmetry evaluation and feedback in normal and trans-tibial amputee subjects. Prosthetics & Orthotics International, 1996. 20(2): p. 101-10. Eng, J.J. and D.A. Winter, Kinetic analysis of the lower limbs during walking: what information can be gained from a three-dimensional model? Journal of Biomechanics, 1995. 28(6): p. 753-8. Blaya, J. and H. Herr, Adaptive Control of a Variable-Impedance Ankle-Foot Orthosis to Assist Drop Foot Gait. IEEE Transactions on Rehabilitation Engineering, 2004. 12(1): p. 24-31. Kuster, M., S. Sakurai, and G.A. Wood, Kinematic and kinetic comparison of downhill and level walking. Clinical Biomechanics, 1995. 10(2): p. 79-84. 81

20.

21.

22.

23.

24.

25. 26. 27.

Hamill, J.a.K., K., Biomechanical Basis of Human Movement. 2 ed. 2003, Philadelphia: Lippincott Williams & Wilkins. Inman, V.T., H.J. Ralston, and F. Todd, Human Walking. 1981, Baltimore, MD.: Williams & Wilkins. Yu, B., et al., Reproducibility of the kinematics and kinetics of the lower extremity during normal stair-climbing. Journal of Orthopedic Research, 1997. 15(3): p. 348-52. McFadyen, B.J. and D.A. Winter, An integrated biomechanical analysis of normal stair ascent and descent. Journal of Biomechanics, 1988. 21(9): p. 733-44. Schumway-Cook, A.a.W., M., Motor Control: Theory and Practical Applications. 2 ed. 2001, Philadelphia: Lippincott Williams & Wilkins.

28. 29.

82

CURRICULUM VITAE

DEANNA GATES
Motion Analysis Laboratory Spaulding Rehabilitation Hospital 125 Nashua Street, Room 745 Boston, MA 02114-1198 (617) 573-2725 dhg7y@bu.edu EDUCATION: 2002-2004 Masters of Science, Biomedical Engineering, Boston University Thesis Title: Characterization of Ankle Function During Stair Ascent, Descent and Level Walking for Ankle Prosthetic and Orthotic Design Advisor: Paolo Bonato, Ph.D. Bachelor of Science, Mechanical Engineering, University of Virginia Minor in Biomedical Engineering, With Honors Undergraduate Thesis Title: Porous Monolithic Sol-gel Columns for Capillary Electrochromatography Advisor: Pamela Norris, Ph.D.

1998-2002

WORK EXPERIENCE: 5/03-present Graduate Research Assistant, Spaulding Rehabilitation Hospital, Motion Analysis Laboratory 9/02-12/03 5/01-5/02 5/00 - 9/00 5/99-9/99 Graduate Teaching Fellow, Boston University Courses taught: Thermodynamics, Medical Imaging Undergraduate Research Assistant, University of Virginia, Aerogel Research Lab Undergraduate Research Assistant, University of Virginia, Mechanical Engineering Department Summer Intern, Architect of the Capitol, Washington D.C., Department of Engineering

83

SKILLS: Computer: AutoCAD, MATLAB, MathCad, Microstation, Microsoft Project, MS Word, MS Excel, WordPerfect, Internet, MS-DOS, C++, Windows NT and Windows 2000 operating systems. Motion Analysis: Vicon 512 Motion Capture System, BodyBuilder software, Optotrak, Bioware software, Temec accelerometers, MotionLabs electromyography, Kistler and AMTI force platforms Related Coursework: Advanced Biomechanics and Kinesiology, Instrumentation for Human Movement Analysis, Gross Human Anatomy, Scientific Basis of Human Movement, Statics, Dynamics, Material Science, and Biomaterials HONORS, AWARDS, AND ORGANIZATIONS: 2002-2003 1998-2002 2001 2000-2002 Awarded Boston University Graduate Teaching Fellowship, 2002-2003 Robert C. Byrd Foundation Scholarship Recipient, 1998 2002 Tau Beta Pi, Engineering Honor Fraternity Member Pi Tau Sigma, Mechanical Engineering Honor Fraternity, Vice President of Membership

84